1. Field of the Invention
The field of the invention is optical waveguides and more particularly, optical waveguide crossovers.
2. Background
In many planar lightwave circuit (PLC) designs, waveguide intersections (or crossovers) are unavoidable. This is particularly true of designs that involve switch interconnect patterns. For instance, for an N×N Spanke switch architecture, there are as many as (N−1)2 crossovers in just one optical path. For example, the maximum number of crossovers in an 8×8 Spanke switch is 49. Each crossover is notorious for contributing to optical loss and can be the source of crosstalk into other channels.
FIG. 1 illustrates an example of a prior art 4×4 Spanke switch. It consists of four 1×4 switch elements (1)–(4) opposite four additional 1×4 switch elements (5)–(8). Each of the four outputs from switch element (1) must have a waveguide connection to a single input on each of switch elements (5), (6), (7), and (8). Likewise, each of the outputs from switch element (2) must have a waveguide connection to a single input on each of switch elements (5), (6), (7), and (8); and so on. This results in an interconnect pattern with waveguide paths having as few as zero crossovers (2 paths), and as many as 9 crossovers (2 paths). For example, connecting waveguide (10) to waveguide (20) requires crossovers at points (31)–(39). There are many ways to realize this interconnect pattern, but as long as it is done in a single plane, these are the minimum number of crossovers.
The usual prior art approach to creating crossovers with minimum optical loss and minimum crosstalk is to design the waveguide pattern such that all waveguide cores intersect at right angles (as shown in FIG. 1), and yet otherwise remain in the same plane. FIG. 2 is a detailed view of the prior art waveguide cores of FIG. 1 at locations (31)–(33). It shows waveguide core (10) occupying the same space as waveguide cores (11), (12), and (13) at locations (31), (32), and (33), respectively.
Because the waveguide cores intersect at right angles, the optical loss and crosstalk are minimized by virtue of the intersecting waveguide (11), (12), or (13) having the minimum vectorial component in the direction of light propagation (25). However, there is still some finite loss caused by each core intersection. This loss arises from diffraction and mode mismatch at each intersection. FIG. 3 is a graphic of the Beam Propagation Method (BPM) simulation results from intersecting waveguide cores. It shows the light propagating in waveguide core (10) as it crosses waveguide core (11). The direction of propagation is shown by the arrow (25). As soon as the light reaches the leading edge of the intersecting waveguide (26), the light is unguided in the x direction and diffracts according to optical diffraction principles. When this diffracted light reaches the opposite side of the intersecting waveguide (27), the E-field profile, or mode profile, is spread out and no longer has the same profile it had when it was originally guided. Therefore, the light will not completely re-couple back into waveguide (10). A fraction of the light will be lost to the cladding as shown (28).
The loss from each crossover can be approximated by:Lcross≈−10·log[1−(4D/v2)(a/w0)4]dB                where:D=(n12−n02)/(2n12)v=(2pa/I)(n12−n02)1/2        
where n0 is the cladding index, n1 is the core index, a is the core half-width, I is the wavelength of the light, and w0 is the radius of the propagating mode at which the E-field is e−1=36.8% of its maximum, E0. It is determined by first evaluating the E-field for points along the radial distance, x, which cannot be solved by closed-form equations:
                                          E            ⁢                                                  ⁢            y                    =                    ⁢                      E0            ·                          cos              ⁡                              (                                  u                  ⁢                                                                          ⁢                                      x                    /                    a                                                  )                                                    ⁢                                                                for          ⁢                                          ⁢                                  x                                      ≤        a                                =                ⁢                  E0          ·                      cos            ⁡                          (              u              )                                ·                      exp            ⁡                          [                                                -                                      (                                          w                      /                      a                                        )                                                  ⁢                                  (                                                                                  x                                                              -                    a                                    )                                            ]                                                                    for          ⁢                                          ⁢                                  x                                      >        a                            where:w=u·tan(u)u=(v2−w2)1/2        
These last two equations must be solved by recursion.
As an example of the loss that can be expected, a waveguide system with the following characteristics:                n0=1.450;        n1=1.482;        a=1.60 um;        I=1.55 um        
will have the following parameters:                D=0.0214;        v=1.986667 radians;        w=1.700426 radians;        u=1.027325 radians.        
w0 is determined by numerically evaluating Ey for several values of x, and finding the value of x where Ey=0.368E0. For this example, this value is w0=1.925 um. Therefore, the loss per crossover (Lcross) is calculated to be approximately 0.045 dB. This result is also obtained by BPM software.
Therefore, there is a need for an improved waveguide crossover that has a lower loss and a method of creating an improved waveguide crossover.